Destination-Initiated RDMA Streaming for Direct Memory Backhaul

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing technologies face challenges in efficiently processing high-throughput, low-latency data streams from remote sources, particularly due to limitations in scalability and resource requirements for RDMA protocol support, and existing solutions like PCIe cards and UDP over networks have limitations in throughput and latency.

Innovation Solution

A host device initiates a direct-to-memory RDMA transmission with a remote device using existing Ethernet or InfiniBand connections, allowing for scalable and efficient processing of streaming data by transitioning into a Ready-to-Receive state and sending instructions to the remote device to enter a Ready-to-Send state, enabling direct data transfer to allocated memory.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If RDMA protocol is used to achieve high-throughput and low-latency data transfer, then throughput and latency performance is improved, but device complexity increases due to the need for full RDMA stack capabilities on both ends

Engineering Contradiction:
Improvedata transfer throughputVSAvoidRDMA stack requirements
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent extracts the RDMA stack capability requirement from the remote device side. The host device initiates RDMA connections and manages the complete RDMA stack locally, while the remote device only needs basic network interface capabilities to send data. This extraction allows high-throughput data transfer without requiring full RDMA stack on all devices.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The host device acts as an intermediary that handles the complex RDMA protocol operations. By centralizing the RDMA stack in the host and using destination-initiated connections, the system eliminates the need for remote devices to implement full RDMA capabilities while maintaining high performance data transfer.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Device complexity

If destination-initiated RDMA connections are used to simplify remote device requirements, then device complexity is reduced, but connection establishment overhead may increase

Engineering Contradiction:
Improveremote device RDMA requirementsVSAvoidconnection establishment time
Core Design Contradiction:
Device complexityVSLoss of time

Solution Approach 1:

The host device performs preliminary actions by pre-configuring memory regions, queues, and connection parameters before actual data transfer begins. The destination device sends initialization data containing connection information, and the host pre-establishes the necessary RDMA infrastructure, allowing for rapid data transfer initiation without time-consuming connection setup during data streaming.

Inventive Principle:
Principle #10Preliminary action

3Productivity

If PCIe cards with FPGAs are used for data processing, then processing capability is improved, but scalability is limited

Engineering Contradiction:
Improvedata processing capabilityVSAvoidsystem scalability
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal data transfer mechanism that works over standard Ethernet or InfiniBand networks. The destination-initiated RDMA approach can be implemented on various platforms including servers, workstations, and cloud instances, making the system adaptable to different hardware configurations and scalable across multiple devices without being tied to specific PCIe card or FPGA implementations.

Inventive Principle:
Principle #6Universality (Multi-functionality)

4Adaptability or versatility

If UDP protocol is used for network data transfer, then scalability is improved by sending additional streams over additional connections, but throughput and latency performance deteriorates

Engineering Contradiction:
Improveconnection scalabilityVSAvoiddata transfer throughput
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent merges the scalability benefits of connection-based protocols with the performance advantages of RDMA. By using destination-initiated RDMA connections, the system can establish multiple parallel connections to different remote devices simultaneously, achieving both scalability and high throughput performance that neither UDP nor traditional RDMA alone could provide.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4687044A1Destination initiated network transmission for data streams
Publication Date: 2026.02.04 NVIDIA CORP
  • EP4687044A1 patent drawingFigure 1
  • EP4687044A1 patent drawingFigure 2
  • EP4687044A1 patent drawingFigure 3

AI summary

Approaches presented herein provide for the real-time backhauling of data streams direct to memory. In at least one embodiment, a host device can initiate a Remote Direct Memory Access (RDMA) transmission with a remote device for sensor data streams. Initialization data for the RDMA transmission can be sent from the host to the remote device, such as over an existing physical Ethernet or InfiniBand (IB) connection. The host device can enter a Ready-to-Receive (RTR) state and send instructions to the remote device to enter a Ready-to-Send (RTS) state. The RDMA transmission can be initialized between the host device and the remote device to transfer streaming data. Using a remote data manager, the streaming data can be sent over the existing Ethernet connection as a RDMA transmission and written directly to allocated memory. The system may be scaled to include higher throughput sensors, additional sensors, and multiple remote transmission devices.